Supplementary Figures

Figure S1

Manhattan plots for two single-trait admixed mappings of ancestry-by-smoking interactions, where smoking is the current smoking exposure (current smokers current non-smokers).

Figure S2

Manhattan plots for two single-trait admixed mappings of ancestry-by-smoking interactions, where smoking is the current heavy smoking exposure (current heavy smokers current moderate smokers and current non-smokers).

Figure S3

Quantile-quantile (QQ) plots for single-trait admixed mapping of ancestry-by-smoking interactions, where smoking is the current smoking exposure (current smokers current non-smokers).

Figure S4

Quantile-quantile (QQ) plots for single-trait admixed mapping of ancestry-by-smoking interactions, where smoking is the current heavy smoking exposure (current heavy smokers current moderate smokers and current non-smokers).

Figure S5

Quantile-quantile (QQ) plots for two multi-trait admixed mappings of ancestry-by-smoking interactions, where smoking is (a) the current smoking exposure (current smokers current non-smokers); or (b) the current heavy smoking exposure (current heavy smokers current moderate smokers and current non-smokers).

Figure S6

Quantile-quantile (QQ) plots for single-trait admixed mapping ({}) of ancestry-by-smoking interactions, where smoking is the current smoking exposure (current smokers current non-smokers). The linear mixed model () (the main text, Section ) developed for admixed mapping was examined by combining three random effects, the two genetic effects \(u_m\) and \(u_i\), and the heterogeneity effects \(u_h\): (a) all \(u_m\), \(u_i\) and \(u_h\) are omitted; (b) only \(u_m\) is included; (c) only two \(u_m\) and \(u_i\) are included; (d) only \(u_h\) is included; (e) only two \(u_m\) and \(u_h\) are included; and (f) all three \(u_m\), \(u_i\) and \(u_h\) are included. The distribution of the test statistic is not inflated only when all the three random effects are presented in the model (the panel (f); \(\lambda\) = 1.06).

Figure S7

Quantile-quantile (QQ) plots for single-trait admixed mapping ({}) (no ancestry-by-smoking interaction). The linear mixed model () (Supplementary Note 1) developed for admixed mapping was examined by combining two random effects, the genetic effects \(u_m\) and the heterogeneity effects \(u_h\): (a) all \(u_m\) and \(u_h\) are omitted; (b) only \(u_m\) is included; (c) only \(u_h\) is included; (d) all two \(u_m\) and \(u_h\) are included. The distribution of the test statistic is not inflated only when the genetic random effects are presented in the model (the panels (b) and (d); \(\lambda\) = 1.04, 1.03). For the later case on panel (d), the \(\chi^2\) statistic was attenuated (the mean of top statistics with P< 0.001 increased from 11.95 to 12.41), because the less amount of residual variance left after modeling the heterogeneity.

Figure S8

Results of fine-mapping analysis in locus chr11:12,000,000-13,000,000 obtained from multi-trait admixed mapping of ancestry-by-smoking interactions, where smoking is the current smoking exposure (current smokers current non-smokers) (Figure (a), the main text). Panel (a) shows the p-values of SNPs (black points) and the names of top 5 associated SNPs; the p-values of long ancestry segments from the upstream admixed mapping analysis, multi-trait (light red color) and single-trait {} (dark red color).

Figure S9

Results of fine-mapping analysis in locus chr11:94,000,000-95,000,000 obtained from multi-trait admixed mapping of ancestry-by-smoking interactions, where smoking smoking is the current heavy smoking exposure (current heavy smokers current moderate smokers and current non-smokers) (Figure (b), the main text). Panel (a) shows the p-values of SNPs (black points) and the names of top 5 associated SNPs; the p-values of long ancestry segments from the upstream admixed mapping analysis, multi-trait (light blue color) and single-trait {} (dark blue color).

Figure S10

Results of fine-mapping analysis in locus chr2:237,750,000-23,9000,000 obtained from single-trait admixed mapping of ancestry-by-smoking interactions, where smoking smoking is the current heavy smoking exposure (current heavy smokers current moderate smokers and current non-smokers) (Figure (d)). Panel (a) shows the p-values of SNPs (black points) and the names of top 5 associated SNPs; the p-values of long ancestry segments from the upstream admixed mapping analysis, multi-trait (light blue color) and single-trait {} (dark blue color).

Figure S11

Distribution of the proportion of global African ancestry among 3,300 African American COPDGene participants. The values ranged from 26.3% to 99.8% and h the mean value of 80.3%.

Figure S12

Distribution of {} trait among three smoking groups: not current smokers; moderate current smokers with 1-14 cigarettes per day; and heavy current smokers with >14 cigarettes per day.

Figure S13

The local ancestry (long segments) averaged across all individuals for Chromosomes (a) 2, (b) 4, (c) 6 and (d) 11. The ancestry inferred using the Hapmap (60 CEU, 60 YRI) is depicted by red lines, and the ancestry inferred using the 1,000 Genomes (99 CEU, 108 YRI) is depicted by blue lines. See also Table for comparison.

Figure S14

Comparison of the global African ancestry inferred using the Hapmap (60 CEU, 60 YRI) and the 1,000 Genomes (99 CEU, 108 YRI) reference panels.

Figure S15

Variance-covariance matrices of the random effects in the linear mixed model () (the main text, Section ) developed for admixed mapping of gene-by-environment interactions: (a) the genetic effects with ARM; (b) the genetic effects with EARM; (c) the medical center effects; (d) the heterogeneity effects; and (e) the residual effects. The relationships are depicted for a subset of 18 individuals, 6 in each of three smoking groups (not current smokers; moderate current smokers with 1-14 cigarettes per day; and heavy current smokers with >14 cigarettes per day). The first 9 individuals have the largest global ancestry proportions, while the last 9 individuals have the smallest global ancestry proportions.